Eliciting Activity Requirements from Crowd Using Genetic Algorithm

Web-based software systems face a wide range of users and situates in different context. Developing such systems needs to deal with the diversity and variability of requirements. Crowd-based requirements engineering performs requirements engineering activities, such as elicitation requirements from the crowd of stakeholders. That leads to the collected requirements being more diverse and wider coverage. However, the requirements elicited from crowd are not directly available and need to be merged into system requirements. It is a tedious and error-prone work without the help of automatic method. System requirements can be expressed in a variety of ways, of which activity diagram is widely used. This paper provides a method based on genetic algorithm. This approach targets to solve two key issues about the individual requirements representation and the requirements synthesis, one is using a triangular matrix encoding scheme to ensure completeness and uniqueness of genetic representation of solution, the other is proposing a generalized information entropy as fitness function to measure candidate solutions. A simple but meaningful example has been used to demonstrate the feasible of this approach. Moreover, during the synthesis of activity diagrams, the information source’s IDs are kept. This can be used for building the traceability links between the system requirements and their source. That will be helpful to requirements management and evolution.